Crowd Monitoring by Video Surveillance System Through Real-Time Object Detection and Tracking Using AI Techniques

Abstract

Real-time video surveillance is an essential component of the modern security system, particularly in the context of monitoring public areas, particularly in densely populated areas such as airports, train stations, and urban junctions. Occlusion, quick movement, and high item density are some of the issues that traditional surveillance systems face when accurately detecting and effectively monitoring objects. YOLO, which stands for "You Only Look Once," is a framework that includes DeepSORT, which stands for "Simple Online and Realtime Tracking with a Deep Association Metric," for effective object tracking. This framework is proposed in this study as a robust multi-object identification and tracking system. By utilizing appearance traits and Kalman filtering, YOLO guarantees that objects are localized and classified quickly and accurately, whereas DeepSORT ensures that identification is maintained consistently between frames. Under various lighting and crowd density situations, the combined model was implemented and tested on video feeds captured from the real world. Given that the system can handle occlusions and re-identify lost targets with high accuracy, low latency, and resilience, as demonstrated by the results of the experiments, it is a feasible option for applications that include intelligent surveillance. When applied in dynamic and complex public contexts, this technique dramatically improves situational awareness and assists proactive security management.